The proposed research would attempt to integrate what are known as the "molar" and "molecular" analyses of operant behavior. The attempted integration would employ a variety of methods, several of which would consist of the development of paradigms to precisely control temporal distributions of responding, that is behavioral patterns, that function as integrated behavioral units, i.e., as operants. The identification of laws governing the establishment and maintenance of such behavioral patterns would help us to understand the boundary conditions within which traditional molar analyses are appropriate and outside of which they are likely to be less than optimal, or perhaps even fundamentally incorrect. The methods proposed to elucidate these laws governing behavioral patterns include the following: the exploration of reinforcement contingencies having the property that not all the behavior on which the delivery of a reinforcer depends is temporally contiguous with reinforcement; the identification of preference relations for different sets of reinforced behavioral patterns; and the description of short-term memory for different behavioral patterns. It is proposed to develop a quantitative learning theory to unify the results on reinforced behavioral patterns, short-term memory, and various "molar" phenomena.